OpenBB-finance/OpenBB · error · ValueError

All columns must be numeric

Error message

All columns must be numeric

What it means

Thrown by the OLS regression endpoint in openbb_econometrics when the selected x_columns/y_column cannot be cast to float via DataFrame.astype(float). statsmodels OLS requires fully numeric design and response matrices, so any non-numeric (string, categorical, date) or NaN-adjacent content in the chosen columns triggers this ValueError, chained from the original pandas cast error.

Source

Thrown at openbb_platform/extensions/econometrics/openbb_econometrics/econometrics_router.py:256

        OBBject with the results being summary object.
    """
    # pylint: disable=import-outside-toplevel
    import re  # noqa
    import statsmodels.api as sm  # noqa
    from openbb_core.app.utils import (
        basemodel_to_df,
        get_target_column,
        get_target_columns,
    )

    X = sm.add_constant(get_target_columns(basemodel_to_df(data), x_columns))
    y = get_target_column(basemodel_to_df(data), y_column)

    try:
        X = X.astype(float)
        y = y.astype(float)
    except ValueError as exc:
        raise ValueError("All columns must be numeric") from exc

    results = sm.OLS(y, X).fit()
    results_summary = results.summary()
    results = {}

    for item in results_summary.tables[0].data:
        results[item[0].strip()] = item[1].strip()
        results[item[2].strip()] = str(item[3]).strip()

    table_1 = results_summary.tables[1]
    headers = table_1.data[0]  # Assuming the headers are in the first row
    for i, row in enumerate(table_1.data):
        if i == 0:  # Skipping the header row
            continue
        for j, cell in enumerate(row):
            if j == 0:  # Skipping the row index
                continue
            key = f"{row[0].strip()}_{headers[j].strip()}"  # Combining row index and column header

View on GitHub (pinned to 3e071fcc2c)

Solutions

1. Verify dtypes before the call: df.dtypes — only pass float/int columns in x_columns and y_column.
2. Coerce the source data: df[c] = pd.to_numeric(df[c], errors='coerce') and dropna() before running the regression.
3. Double-check the column names in x_columns/y_column against the actual dataset columns (get_target_column also fails loudly on missing names).
4. If a categorical regressor is intended, encode it (dummies) first.

Example fix

# before
res = obb.econometrics.ols(data, y_column='revenue', x_columns=['sector', 'growth'])  # sector is a string

# after
df = data.to_df()
X = pd.get_dummies(df[['sector']], drop_first=True)
df = pd.concat([df[['revenue', 'growth']].apply(pd.to_numeric, errors='coerce'), X], axis=1).dropna()
res = obb.econometrics.ols(Data(data=df), y_column='revenue', x_columns=['growth', 'sector_technology'])

df = data.to_df()
cols = [y_column] + list(x_columns)
assert all(c in df.columns for c in cols), 'missing column(s)'
non_numeric = [c for c in cols if not pd.api.types.is_numeric_dtype(df[c])]
assert not non_numeric, f'non-numeric columns: {non_numeric}'
df = df[cols].apply(pd.to_numeric, errors='coerce').dropna()

def columns_are_numeric(df, columns: list[str]) -> bool:
    """True when every named column exists and has a numeric dtype."""
    return all(c in df.columns and pd.api.types.is_numeric_dtype(df[c]) for c in columns)

try:
    res = obb.econometrics.ols(data, y_column=y, x_columns=xs)
except ValueError as e:
    if str(e) == 'All columns must be numeric':
        # coerce and retry
        ...

When it happens

Trigger: Calling obb.econometrics.ols() (or the regression router command at line ~256) with y_column or an entry in x_columns referring to a string/categorical/date column; passing a dataset where numeric columns are stored as object dtype after JSON round-tripping.

Common situations: Loading data from CSV/JSON where numbers arrive as strings, selecting a date or symbol column as a regressor by mistake, or provider data whose schema changed a field from float to string between versions.

Related errors

• This analysis requires at least 3 items in the dataset.
• Calculation asks for at least last {window} days of data
• Error: expected data with numeric values.

AI-assisted analysis of OpenBB-finance/OpenBB@3e071fcc2c (2026-08-14). Data as JSON: /api/errors/4c431f3057449a0d. Report an issue: GitHub.